Method and system for monitoring limulus population in intertidal zone based on unmanned aerial vehicle remote sensing image

Abstract

The application provides a kind of intertidal zone hatching population monitoring method and system based on unmanned aerial vehicle remote sensing image, the method comprises: according to the topographic information and tidal law information of target intertidal zone area, in the target intertidal zone area, several sample lines parallel to corresponding coast are preset, and several investigation quadrat areas are arranged on the sample line;In the preset tidal time window, control unmanned aerial vehicle to carry out aerial survey flight to each investigation quadrat area, obtain unmanned aerial vehicle remote sensing image set;The unmanned aerial vehicle remote sensing image in the unmanned aerial vehicle remote sensing image set is respectively preprocessed and image optimized, and the orthographic image data to be detected is obtained;The orthographic image data is input into the preset hatching population identification detection model, target detection and identification are carried out, and the hatching population detection result image of the target intertidal zone area is obtained. Therefore, by applying the technical solution of the application, the detection accuracy and monitoring efficiency of intertidal zone hatching population can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of geographic information technology, and in particular to a method and system for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery. Background Technology

[0002] The intertidal horseshoe crab population is an important ecological indicator species, and monitoring its population dynamics is of great significance for biodiversity conservation and coastal ecological management.

[0003] Traditional monitoring of horseshoe crabs in the intertidal zone relies primarily on manual field surveys. Surveyors must visually count and measure the horseshoe crab population within a limited time window after low tide. This method is not only inefficient and has limited coverage, but it is also constrained by tides, weather, and complex terrain, making it difficult to achieve large-scale, high-frequency continuous monitoring. Furthermore, manual surveys are easily influenced by subjective experience, resulting in low detection rates of small targets or concealed individuals such as juvenile horseshoe crabs, and compromising data accuracy and consistency.

[0004] In recent years, UAV remote sensing technology has been introduced into the field of ecological monitoring due to its flexibility and efficiency. However, existing detection and identification methods based on UAV imagery mostly use general target detection models and aerial survey methods. They have not been optimized for the biological characteristics and habitat of intertidal horseshoe crabs (especially juvenile horseshoe crabs) in terms of feature extraction, model architecture, and survey process. They are not well adapted to the characteristics of horseshoe crabs in the intertidal background, such as small size, variable morphology, high similarity to the background, and partial occlusion. This results in limited detection accuracy and makes it difficult to meet the needs of accurate and efficient population monitoring. Summary of the Invention

[0005] Therefore, the purpose of this application is to provide a method and system for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery, which can effectively improve the detection accuracy and monitoring efficiency of intertidal horseshoe crab populations.

[0006] The objective of this application can be achieved through the following technical solutions:

[0007] A method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery includes the following steps: Based on the topographic and tidal information of the target intertidal region, several transects parallel to the corresponding coastline are pre-defined in the target intertidal region, and several survey quadrats are established along these transects; within a pre-defined tidal time window, a UAV is controlled to conduct aerial survey flights over each of the survey quadrats to obtain a set of UAV remote sensing images, wherein the set of UAV remote sensing images includes UAV remote sensing images corresponding to each of the survey quadrats; the UAV remote sensing images in the set of UAV remote sensing images are pre-processed and optimized to obtain orthophoto data to be detected; the orthophoto data is input into a pre-defined horseshoe crab population identification and detection model for target detection and recognition to obtain a horseshoe crab population detection result image of the target intertidal region, wherein the horseshoe crab population detection result image is marked with the location identifiers of several horseshoe crab targets.

[0008] A monitoring system for intertidal horseshoe crab populations based on UAV remote sensing imagery, characterized by comprising: a survey quadrat area setting module, used to pre-define several transects parallel to the corresponding coastline in the target intertidal region according to the topographic information and tidal pattern information of the target intertidal region, and to set up several survey quadrat areas on the transects; and a UAV remote sensing image set acquisition module, used to control a UAV to conduct aerial survey flights over each of the survey quadrat areas within a preset tidal time window to obtain a UAV remote sensing image set, wherein the UAV remote sensing images The dataset includes UAV remote sensing images corresponding to each of the survey sample plots; an orthophoto data acquisition module, used to preprocess and optimize the UAV remote sensing images in the dataset to obtain orthophoto data to be detected; and a horseshoe crab population detection result image acquisition module, used to input the orthophoto data into a preset horseshoe crab population identification and detection model to perform target detection and recognition, and obtain horseshoe crab population detection result images of the target intertidal zone, wherein the horseshoe crab population detection result images are marked with the location identifiers of several horseshoe crab targets.

[0009] Compared to existing technologies, the method described in this application firstly constructs a standardized spatial sampling framework that conforms to the vertical distribution characteristics of horseshoe crab populations by pre-setting transects and survey quadrats based on the topography and tidal patterns of the target intertidal zone, thus avoiding monitoring errors caused by the arbitrary deployment of traditional manual sampling methods. Secondly, it controls UAVs to conduct aerial surveys within a preset tidal time window, ensuring consistency in the exposure status and lighting conditions of the intertidal mudflats during data collection. Next, it performs preprocessing procedures, including geometric stitching, distortion correction, and radiometric optimization, on the acquired UAV remote sensing images to generate orthophoto data with unified geographic coordinates and clear texture. Finally, it inputs the orthophoto data into a preset horseshoe crab population identification and detection model, realizing automated detection and location marking of horseshoe crab targets. Therefore, the method described in this application combines a standardized habitat-adapted sampling scheme, time-constrained aerial survey data acquisition, a targeted remote sensing image processing workflow, and a dedicated horseshoe crab population identification and detection model to form a complete automated monitoring technology system for intertidal horseshoe crab populations. This significantly improves the detection accuracy and monitoring efficiency of intertidal horseshoe crab populations while expanding the detection range and ensuring data timeliness.

[0010] To better understand and implement this application, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0011] Figure 1 A flowchart illustrating a method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery provided in this application; Figure 2 A flowchart illustrating the steps for acquiring orthophoto data in the method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery provided in this application; Figure 3 A flowchart illustrating the steps for species classification of horseshoe crabs in the intertidal horseshoe crab population monitoring method based on UAV remote sensing imagery provided in this application; Figure 4 A flowchart illustrating the steps for obtaining individual biological parameters of horseshoe crabs in the intertidal horseshoe crab population monitoring method based on UAV remote sensing imagery provided in this application; Figure 5 This application provides a structural principle diagram of an intertidal horseshoe crab population monitoring system based on UAV remote sensing imagery. Detailed Implementation

[0012] This application provides a method and system for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery. To make the objectives, technical solutions, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0013] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0014] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0015] The invention will be further explained below with reference to the accompanying drawings and the description of the embodiments.

[0016] Example 1 Please refer to Figure 1 , Figure 1 A flowchart illustrating a method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery, provided in this application. The method includes the following steps: S10: Based on the topographic information and tidal pattern information of the target intertidal zone, several transects parallel to the corresponding coastline are pre-set in the target intertidal zone, and several survey quadrat areas are set up on the transects. S20: Within the preset tidal time window, control the UAV to conduct aerial survey flights over each of the survey sample areas to obtain a set of UAV remote sensing images; S30: Perform preprocessing and image optimization on the UAV remote sensing images in the UAV remote sensing image set to obtain orthophoto data to be detected. S40: Input the orthophoto data into the preset horseshoe crab population identification and detection model to perform target detection and identification, and obtain the horseshoe crab population detection result image of the target intertidal region.

[0017] Compared to existing technologies, the technical solution of this application addresses the technical challenges in monitoring intertidal horseshoe crab (especially juveniles) populations, including small target scale, complex background, arbitrary posture, difficulty in species differentiation, and lack of individual information. It constructs a complete and automated solution from three levels: standardized survey procedures, specialized model architecture, and precise information inversion. Specifically: Regarding the standardized survey procedure, this application ensures the scientific validity and comparability of the sampling space by pre-setting transects and quadrats based on topography and tidal patterns; it guarantees the quality and consistency of input data by conducting UAV aerial surveys within a precise time window after low tide and setting optimized aerial photography parameters; and in terms of the horseshoe crab population identification and detection model architecture, this application integrates high-frequency and low-frequency feature enhancement modules into the corresponding feature extraction network to respectively enhance the identification of horseshoe crab target details and overall characteristics. The application enhances semantic capture capabilities by fusing the aforementioned features through a dedicated module in the corresponding feature fusion network. The detection method is then configured as a rotated rectangular bounding box detection, supplemented by a hybrid loss function including rotation angle loss during training. This significantly improves the localization and recognition accuracy of horseshoe crabs in any posture against the complex intertidal mudflat background. Furthermore, following the detection head of the horseshoe crab population identification and detection model, this application adds a species classification sub-network driven by horseshoe crab morphological feature vectors, enabling automatic differentiation between Chinese horseshoe crabs and roundtail horseshoe crabs. Moreover, in terms of precise information inversion, this application, based on the parameters of the rotated rectangular bounding box and the ground resolution of the corresponding orthophoto data, achieves accurate measurement of the pre-carapace width and other dimensions of individual horseshoe crab targets. A dedicated "pre-carapace width-age" nonlinear regression model is constructed to invert the age of horseshoe crab targets (especially juveniles).

[0018] Therefore, the technical solution of this application, by organically combining a standardized survey design for habitat adaptation, a targeted image acquisition and processing process, and a dedicated detection-classification-inversion model that deeply integrates morphological prior knowledge, achieves full-chain innovation from data collection to population parameter extraction. This significantly improves the monitoring efficiency, data accuracy, and information dimensions (species, quantity, size, age) of intertidal horseshoe crab populations, and solves the limitations of existing technologies in the monitoring of intertidal horseshoe crab populations.

[0019] For step S10: Based on the topographic information and tidal pattern information of the target intertidal zone, several transects parallel to the corresponding coastline are pre-set in the target intertidal zone, and several survey quadrat areas are set up on the transects.

[0020] The target intertidal zone is a specific tidal flat area located at the junction of the ocean and land, where horseshoe crab population monitoring is to be conducted. This target intertidal zone is submerged at high tide and exposed at low tide. In this application, the target intertidal zone specifically refers to tidal flat areas where horseshoe crabs may inhabit. The topographic information includes, but is not limited to, data describing the surface morphology of the target intertidal zone, such as the coastline orientation, tidal flat slope, width, and micro-topography (e.g., tidal channels, silty beaches). The tidal pattern information refers to the type of tide (e.g., semi-diurnal tide, diurnal tide), tide height, tidal rise and fall times, and the periodicity of spring tides and neap tides in the sea area corresponding to the target intertidal zone. The transect is a virtual reference line parallel to the coastline, pre-defined for planning the survey area, used to systematically organize sampling units. The survey quadrat is a standard area unit set on the transect for quantitative surveys. In this application, the survey quadrat is the basic spatial unit for monitoring horseshoe crab populations.

[0021] In one embodiment, step S10 aims to establish a standardized, repeatable spatial sampling framework that fits the habitat characteristics of horseshoe crabs for UAV aerial surveys and subsequent analysis. Its core is to pre-determine the location of survey units according to scientific principles, in order to replace the traditional random or empirical deployment method in the field, thereby improving the standardization of horseshoe crab population monitoring and the comparability of the corresponding data.

[0022] In one embodiment, step S10 includes: S101: Based on the topographic information and tidal pattern information, determine the seawater inundation frequency distribution of the target intertidal zone.

[0023] The seawater inundation frequency distribution is a spatial gradient model describing the probability or frequency of seawater inundation at different locations in the intertidal zone. It integrates topographic elevation and tidal level change data to calculate the proportion or frequency of time that different areas are covered by seawater in a unit of time (such as one tidal cycle or one month) from the shore to the sea. In this application, the seawater inundation frequency distribution is used to objectively quantify the inundation environment gradient of the target intertidal zone area, serving as a scientific basis for dividing ecological zones and laying transects.

[0024] S102: Based on the seawater inundation frequency distribution, three transects are pre-defined parallel to the corresponding coastline.

[0025] The three transects are arranged sequentially in a direction perpendicular to the coastline and correspond to three preset gradient threshold intervals from low to high in the seawater inundation frequency distribution. The preset gradient threshold intervals are three continuous intervals artificially or naturally divided according to the numerical range of seawater inundation frequency, representing three typical habitats with low, medium, and high inundation frequencies. In this application, these three continuous intervals correspond to the ecological high tide zone, mid-tide zone, and low tide zone, respectively. The high tide zone has the lowest inundation frequency and the daily inundation time is short; the low tide zone has the highest inundation frequency and the daily inundation time is long; and the mid-tide zone is in between.

[0026] In this embodiment, by pre-setting three transects located in these three intervals, it can be ensured that the sampling covers all major habitat types that horseshoe crab populations (especially juvenile horseshoe crabs) may inhabit, thereby improving the detection accuracy.

[0027] S103: On each of the aforementioned transects, 3 to 6 survey quadrats with a size of 10 meters × 10 meters are arranged at equal intervals.

[0028] Wherein, on the same transect line, the distance between any two adjacent survey quadrat areas is 100 meters; of course, in some embodiments, the distance can also be set between 100 meters and 300 meters, and those skilled in the art can select and adjust it according to the actual situation.

[0029] In one specific embodiment, step S10 is implemented as follows: First, digital elevation model (DEM) data of the target intertidal zone and local annual tide forecast data are collected. These are then overlaid and analyzed using Geographic Information System (GIS) software to simulate and calculate the seawater inundation frequency of each pixel point within a spring tide cycle, generating a frequency distribution map. Next, on this distribution map, three consecutive intervals are delineated from the shore to the sea based on frequency values ​​from low to high (e.g., 0-20% frequency corresponds to high tide, 20-60% to mid-tide, and 60-100% to low tide). Then, in a direction parallel to the coastline, a transect (i.e., three transects) is digitally pre-defined in each of the high tide, mid-tide, and low tide zones. Finally, on each digital transect, a marker point is set every 100 meters, with the transect starting point as the reference. A square surface feature with a side length of 10 meters is generated centered on each marker point until the upper limit of the transect length is reached (usually forming 3-6 survey quadrats).

[0030] For step S20: Within the preset tidal time window, control the UAV to conduct aerial survey flights over each of the survey sample areas to obtain a set of UAV remote sensing images.

[0031] The UAV remote sensing image set includes UAV remote sensing images corresponding to each of the survey sample plot areas.

[0032] In one embodiment, step S20 aims to utilize a drone platform to efficiently and standardly acquire high-quality raw image data covering the entire survey quadrat area according to pre-planned optimal spatiotemporal parameters, transforming traditional manual field surveys into automated, high-precision aerial data acquisition, which is a key data source to ensure the accuracy of subsequent image processing and target identification.

[0033] In one embodiment, step S20 includes: S201: Obtain the coordinate information of the surveyed sample plot area using RTK positioning technology.

[0034] The RTK positioning technology is a real-time dynamic carrier phase differential technology, which is a satellite navigation positioning measurement method that can obtain centimeter-level positioning accuracy in the field through real-time data communication between the base station and the rover (UAV). In this embodiment, the technology is used to accurately obtain the real geographic coordinates (such as latitude, longitude, and elevation) of the center point or corner point of each survey sample area preset in step S10, so as to ensure that subsequent route planning and data collection can be accurately aligned with the target area.

[0035] S202: Generate UAV aerial survey routes based on the coordinate information of the survey sample area.

[0036] The UAV aerial survey route refers to a set of three-dimensional spatial waypoint sequences and flight commands automatically calculated by ground control software based on the boundary coordinates of the task area (i.e., all survey quadrat areas) and preset flight parameters (such as altitude, speed, and overlap rate) to control the UAV to fly automatically and complete the image acquisition task. In this embodiment, the generated UAV aerial survey route will ensure that the UAV can systematically fly over each survey quadrat area and acquire images according to the set parameters.

[0037] S203: Within a preset tidal time window, control the UAV to conduct aerial survey flights according to the UAV aerial survey route, and collect UAV remote sensing images corresponding to each of the survey sample areas to obtain the UAV remote sensing image set. The preset tidal time window refers to a specific tidal state time period selected in advance for conducting intertidal zone surveys. In this embodiment, the preset tidal time window is 0.5 hours after the spring tide recedes. The forward overlap rate is the percentage of the overlap portion along the flight direction between two adjacent UAV remote sensing images on the same flight path, relative to the length of a single image. In this embodiment, the forward overlap rate of the UAV is 80%. The lateral overlap rate is the percentage of the overlap portion perpendicular to the flight direction between two adjacent parallel UAV remote sensing images, relative to the width of a single image. In this application, the lateral overlap rate of the UAV is 70%. The flight altitude refers to the vertical distance between the center of the camera lens carried by the UAV and the local ground when performing aerial surveying tasks. In this application, the flight altitude of the UAV during aerial surveying flights is 5 to 10 meters. The ground resolution is the actual ground size represented by one pixel on the UAV remote sensing image. In this application, the ground resolution of the UAV remote sensing images collected by the UAV is not less than 0.5 centimeters.

[0038] In one specific embodiment, step S20 is implemented as follows: First, the corner coordinates of all survey quadrat areas recorded by RTK measurement are imported into the UAV ground station control software (such as DJI Pilot 2 or similar professional software); then, the aerial survey mission parameters are set in the corresponding software: the flight altitude is set to 8 meters, the forward overlap rate is set to 80%, the lateral overlap rate is set to 70%, and the camera lens is vertically downward; the corresponding software will automatically generate the optimal zigzag automatic flight route according to the distribution range of the survey quadrat areas; then, about 0.5 hours after the pre-determined local spring and low tide times, the UAV is launched in a safe area on the shore and the flight route mission is uploaded; the UAV will automatically take off, fly along the planned route in the flight route mission, and automatically trigger the camera to take pictures at each waypoint until the coverage of all survey quadrat areas is completed; finally, the UAV automatically returns and exports all UAV remote sensing images with accurate POS (position and attitude) data collected from the memory card to form a UAV remote sensing image set.

[0039] For step S30: preprocess and optimize the UAV remote sensing images in the UAV remote sensing image set to obtain the orthophoto data to be detected.

[0040] The preprocessing and image optimization involve performing a series of computer processing operations on the original UAV remote sensing images to eliminate errors, enhance useful information, and form standard format data.

[0041] In one embodiment, step S30 aims to transform the raw, scattered UAV remote sensing images obtained in step S20 into a complete, accurate, and visually optimized regional orthophoto map, thereby solving data quality problems caused by changes in UAV shooting posture, lens distortion, differences in lighting conditions, and inconsistencies in coordinates between multiple images, and providing a reliable and standardized data base for subsequent identification.

[0042] Please refer to Figure 2 , Figure 2 A flowchart illustrating the steps involved in acquiring orthophoto data in the method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery provided in this application.

[0043] In one embodiment, step S30 includes: S301: Perform image stitching, distortion correction, and geographic coordinate alignment on the UAV remote sensing images in the UAV remote sensing image set to generate intermediate orthophoto data.

[0044] The image stitching refers to the process of seamlessly merging multiple independent UAV remote sensing images into a continuous, large-scale digital image covering the entire survey area by utilizing the overlapping areas between adjacent images and through feature point matching and coordinate transformation. The distortion correction is the process of mathematically correcting shape distortions (such as barrel distortion, pincushion distortion, etc.) caused by lens optical characteristics in UAV remote sensing images based on the physical parameter model of the camera lens, restoring the true geometric shape of ground features. The geographic coordinate alignment is the process of assigning precise geographic coordinates (such as the WGS84 coordinate system) to the stitched UAV remote sensing image using POS (Position and Attitude System) data or ground control points recorded by the UAV, so that each pixel on the UAV remote sensing image corresponds to a unique geographic location in the real world.

[0045] S302: The intermediate orthophoto data is subjected to illumination equalization processing using a contrast-limited adaptive histogram equalization method to obtain the orthophoto data to be detected. ; Where I represents the intermediate orthophoto image data, clipLimit is a preset contrast limit coefficient set to 2.0, and tileGridSize is the corresponding tile grid size in the illumination equalization process, which is 8 rows × 8 columns. I clahe The orthophoto data to be detected is CLAHE(), which is a preset equalization function.

[0046] The contrast-limited adaptive histogram equalization method is an image processing algorithm for image enhancement. Unlike global histogram equalization, the contrast-limited adaptive histogram equalization method divides the input image into multiple small context regions (i.e., blocks), performs histogram equalization independently on each block to enhance local contrast, and prevents excessive amplification of noise by limiting contrast. The tileGridSize parameter defines the grid size of the blocks when the contrast-limited adaptive histogram equalization method is executed, that is, the number of blocks the image is divided into in the row and column directions. Setting it to 8 rows × 8 columns is equivalent to dividing the image (the image corresponding to the intermediate orthophoto data) into 8x8=64 local regions for processing to adapt to the illumination changes in different areas of the tidal flat image.

[0047] In this embodiment, step S302 can effectively improve the problem of uneven lighting caused by backlight, cloud shadows or water film reflection on the tidal flats, highlight the texture and outline features of the horseshoe crab target, and improve its visibility and distinguishability against complex backgrounds.

[0048] In one specific embodiment, step S30 is implemented as follows: First, the UAV remote sensing image set and its corresponding POS data obtained in step S20 are imported into professional photogrammetry software (such as Pix4Dmapper or Agisoft Metashape, etc.); the corresponding software automatically executes the following process: aerial triangulation is performed to calculate the precise exterior orientation elements of all UAV remote sensing images; based on the calculation results, lens distortion correction and dense matching of multi-view images are performed simultaneously to generate a 3D point cloud; then, a digital surface model is constructed based on the point cloud, and each corrected image is "pasted" onto the model through orthographic projection, and finally stitched together. First, generate a complete intermediate orthophoto image (DOM) with accurate geographic coordinates. Then, import this intermediate orthophoto image data into an image processing environment (such as a Python script using the OpenCV library), call the cv2.createCLAHE function to create a CLAHE object, set the parameters clipLimit=2.0 and tileGridSize=(8,8), use this object to process each color channel (usually RGB) of the intermediate orthophoto image separately, and merge the processing results to finally obtain high-quality orthophoto image data to be detected after illumination equalization, which is used for horseshoe crab target identification in subsequent steps.

[0049] For step S40: Input the orthophoto data into the preset horseshoe crab population identification and detection model to perform target detection and identification, and obtain the horseshoe crab population detection result image of the target intertidal region.

[0050] The horseshoe crab population detection result image is marked with the location identifiers of several horseshoe crab targets.

[0051] In one embodiment, before the orthophoto data is input into the preset horseshoe crab population identification and detection model, those skilled in the art may further perform pixel cropping and normalization processing on it to adapt to the input requirements of the preset horseshoe crab population identification and detection model.

[0052] In one embodiment, the preset horseshoe crab population identification and detection model is a target detection neural network model based on an improved YOLO architecture; wherein, the improvement includes: the feature extraction network of the horseshoe crab population identification and detection model integrates a high-frequency feature enhancement module for extracting detailed features of the horseshoe crab target and a low-frequency feature enhancement module for extracting the overall semantic features of the horseshoe crab target; the feature fusion network of the horseshoe crab population identification and detection model is provided with a high-low frequency feature fusion module, which is used to fuse features from the high-frequency feature enhancement module and the low-frequency feature enhancement module, and output the fused multi-scale feature map; the detection method of the horseshoe crab population identification and detection model is configured as a rotated rectangular bounding box detection; the detection head of the horseshoe crab population identification and detection model is configured to output the parameters of the corresponding rotated rectangular bounding box, the parameters including the center coordinates, width, height, and rotation angle relative to the horizontal axis of the rotated rectangular bounding box.

[0053] In this embodiment, the high-frequency feature enhancement module processes the input first feature map in the following manner to output an enhanced high-frequency feature map: S41: Perform a 1×1 convolution operation on the first feature map, and multiply the resulting convolution with a preset spatial attention mechanism weight matrix to obtain the high-frequency feature map: ; in, F high The high-frequency feature map, X in For the first feature map, Conv 1×1 (·) represents the convolution function corresponding to a convolution operation with a 1×1 kernel. S attention The preset spatial attention mechanism weight matrix, F high This refers to the high-frequency feature map.

[0054] In this embodiment, the low-frequency feature enhancement module processes the input second feature map in the following manner to output an enhanced low-frequency feature map: S42: Perform a dilated convolution operation with a dilation rate of 2 and a kernel size of 3×3 on the second feature map to obtain the low-frequency feature map: ; in,F low For the low-frequency feature map, X in ’ For the second feature map, AtrousConv 3×3 (·) represents the dilated convolution function corresponding to a dilated convolution operation with a 3×3 kernel. r =2 represents the void ratio. F low This is the low-frequency feature map.

[0055] In this embodiment, the high- and low-frequency feature fusion module fuses the input high-frequency feature map and the low-frequency feature map in the following manner to output the multi-scale feature map: S43: The high-frequency feature map and the low-frequency feature map are concatenated, and the concatenation result is subjected to a preset linear transformation to obtain the multi-scale feature map. ; in, F fusion For the multi-scale feature map, Concat() is the concatenation function corresponding to the concatenation process. W fusion The preset fusion weight matrix, b fusion For the preset bias, F fusion This refers to the multi-scale feature map.

[0056] In one embodiment, those skilled in the art will understand that the first feature map and the second feature map may be the same feature map or different feature maps.

[0057] In one embodiment, before the detection head of the horseshoe crab population identification and detection model outputs the parameters of the corresponding rotated rectangular bounding box, those skilled in the art can also perform non-maximum suppression processing on the corresponding rotated rectangular bounding box to propose repeated target boxes with an overlap rate greater than or equal to 0.3 (i.e., the rotated rectangular bounding boxes in the horseshoe crab population detection result image), thereby improving the readability of the obtained rotated rectangular bounding boxes.

[0058] In one embodiment, the horseshoe crab population identification and detection model is obtained by training a preset minimum mixture loss function; the preset minimum mixture loss function is obtained by weighted summation of the bounding box regression loss, the target classification loss, and the confidence loss, and its form is as follows: ; in, L OBB The preset minimum mixture loss function, Lreg The regression loss of the rotated frame, L cls The classification loss for the target is... L conf For the confidence loss, α The preset first loss weighting coefficient, β The second loss weighting coefficient is preset. γ The third loss weight coefficient is preset, and α =1.5, β =1.0, γ =0.5.

[0059] In this embodiment, the rotation frame regression loss is calculated using the following formula: ; in, L reg The regression loss of the rotated frame, BB pred For the predicted rotated rectangular bounding box, BB gt For the corresponding labeled true rotated rectangular bounding box, GIoU() is the preset generalized intersection-union ratio (IU) calculation function. λ The preset angle loss weight coefficient and λ =0.8, θ pred For the predicted rotation angle, θ gt This corresponds to the actual rotation angle.

[0060] In this embodiment, the target classification loss is calculated using the following formula: ; in, L cls The classification loss for the target is... α t The preset positive and negative sample weight coefficients and α t =0.25, δ is the preset difficult sample mining coefficient and δ=2, P t The predicted classification probability of the horseshoe crab target by the horseshoe crab population identification and detection model is given.

[0061] In this embodiment, the confidence loss is calculated using the following formula: ; in, L conf For the confidence loss, yThis represents the true confidence level label and takes a value of 1 or 0. σ (·) represents the preset sigmoid activation function. p The prediction confidence level of the horseshoe crab population identification and detection model is given.

[0062] In one embodiment, a species classification subnetwork is connected to the detection head of the horseshoe crab population identification and detection model; the species classification subnetwork is used to classify the detected horseshoe crab targets based on the horseshoe crab population detection result image; therefore, this application also provides some monitoring steps that can be used in the intertidal horseshoe crab population monitoring method based on UAV remote sensing imagery to distinguish the species of the detected horseshoe crab targets; please refer to Figure 3 , Figure 3 The flowchart of the steps for species classification of horseshoe crab targets in the intertidal horseshoe crab population monitoring method based on UAV remote sensing imagery provided in this application, after the step of obtaining the horseshoe crab population detection result image of the target intertidal region, the intertidal horseshoe crab population monitoring method further includes: S51: Based on the biological morphological information of the Chinese horseshoe crab, obtain the biological feature vector V of the Chinese horseshoe crab. C : V C =(V C1 V C2 V C3 ); Among them, V C1 V represents the cephalothorax shape characteristics of the Chinese horseshoe crab. C2 V represents the sharpness of the lateral spines of the horseshoe crab described above. C3 The number of lateral spines of the Chinese horseshoe crab.

[0063] S52: Based on the biological morphological information of the horseshoe crab, obtain the biological feature vector V of the horseshoe crab. R : V R =(V R1 V R2 V R3 ); Among them, V R1 V represents the cephalothorax shape characteristics of the horseshoe crab. R2 V represents the sharpness of the lateral spines of the horseshoe crab. R3 The number of lateral spines of the horseshoe crab is given.

[0064] S53: Input the horseshoe crab population detection result image into the species classification sub-network, and drive the species classification sub-network to process the horseshoe crab population detection result image according to the biofeature vectors of the Chinese horseshoe crab and the roundtail horseshoe crab to obtain the species classification probability corresponding to the detected horseshoe crab targets: P spec= (PC, PR); Among them, P spec The probability of species classification corresponding to the detected horseshoe crab target is represented by PC, which indicates the probability that the corresponding horseshoe crab target is the Chinese horseshoe crab, and PR indicates the probability that the corresponding horseshoe crab target is the roundtail horseshoe crab.

[0065] S54: Based on the species classification probability, determine the species of the corresponding horseshoe crab target. If PC is greater than 0.5, the horseshoe crab target is determined to be the Chinese horseshoe crab. If PR is greater than 0.5, the horseshoe crab target is determined to be the roundtail horseshoe crab.

[0066] In one embodiment, the species classification subnetwork includes at least two fully connected layers and a classification output processing layer (i.e., Softmax (·)). After the horseshoe crab population detection result image is input into the species classification subnetwork, the species feature vector is first extracted through the first fully connected layer, the preset ReLU activation function, and the second fully connected layer, and then the corresponding species classification probability is output through the Softmax function.

[0067] In addition, this application also provides some monitoring steps that can be used in the aforementioned method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery, in order to predict the age of the horseshoe crab target; please refer to Figure 4 , Figure 4 The flowchart illustrates the steps for obtaining individual biological parameters of horseshoe crab targets in the intertidal horseshoe crab population monitoring method based on UAV remote sensing imagery provided in this application. Following the step of obtaining the horseshoe crab population detection result image of the target intertidal region, the intertidal horseshoe crab population monitoring method further includes the step of obtaining the corresponding individual biological parameters of the horseshoe crab targets: S61: For the horseshoe crab target in the horseshoe crab population detection result image, based on its corresponding rotated rectangular bounding box, extract the pixel width of its forebody armor using a preset edge detection algorithm. W p : ; The left and right edge pixel coordinates of the horseshoe crab target's forebody are respectively ( x 1, y 1) and ( x 2, y 2) θ The rotation angle of the rotating rectangular bounding box. W p The pixel width is [value].

[0068] S62: Combining the ground resolution of the orthophoto data, and based on the pixel width, calculate the actual physical width of the horseshoe crab's forebody armor: ; in,W real The actual physical width of the forebody armor of the horseshoe crab target. GSD The ground resolution of the orthophoto data is given. k This is the preset unit conversion factor.

[0069] In one embodiment, the preset unit conversion factor is 10, used to convert centimeters to millimeters.

[0070] In one embodiment, those skilled in the art can also exclude falsely detected targets (such as tidal groove textures or silt clumps) based on the width / length ratio of the horseshoe crab's protocarp, the target pixel area, and texture features: if the width / length ratio of the horseshoe crab's protocarp is not in the closed range of 0.8 to 1.2, and / or, the target pixel area of ​​the horseshoe crab is not in the closed range of 5 to 500 (adapting to the size of horseshoe crabs at a ground resolution of 0.5 cm), and / or, the contrast of the horseshoe crab's texture features after processing with a preset grayscale co-occurrence matrix is ​​less than 0.15, then the horseshoe crab is determined to be a falsely detected target.

[0071] S63: If the horseshoe crab's protocarp is occluded in the horseshoe crab population detection result image, then the width of the horseshoe crab's crawl marks is obtained, and the actual physical width of the horseshoe crab's protocarp is calculated using a preset correction formula. W real =1.2× W trace +0.8; in, W trace The width of the crawl marks on the horseshoe crab target.

[0072] S64: Input the actual physical width into the preset age inversion model to obtain the predicted age of the horseshoe crab target: ; in, a and b These are the preset regression coefficients of the first model and the regression coefficients of the second model, respectively. a =4.263, b =-7.892, N pred The predicted age of the horseshoe crab target.

[0073] In one embodiment, those skilled in the art may also round down the predicted age of the horseshoe crab target according to the actual situation.

[0074] In one embodiment, those skilled in the art can calculate the preset first model regression coefficient and second model regression coefficient by statistically analyzing the actual physical width and age of the horseshoe crab precursor carapace. Specifically, after statistical analysis of a large number of samples, those skilled in the art obtained the following statistical data: 1st age - 6 mm, 2nd age - 8 mm, 3rd age - 11 mm, 4th age - 14 mm, 5th age - 18 mm, 6th age - 22 mm, 7th age - 28 mm, 8th age - 37 mm, 9th age - 44 mm, 10th age - 50 mm, 11th age - 70 mm, 12th age - 87 mm, 13th age - 113 mm, 14th age - 147 mm, from which the following can be calculated. a =4.263, b =-7.892.

[0075] In one specific embodiment, the actual physical width of the protocarp of a horseshoe crab was measured to be 48 mm. This value was then input into the preset age inversion model, and N = 4.263 × ln48 was calculated. 7.892 ≈ 4.263 × 3.871 7.892≈9, which is in line with the preset error (0.5 years).

[0076] Example 2 Please refer to Figure 5 This application also provides a horseshoe crab population monitoring system based on UAV remote sensing imagery to implement the steps of the horseshoe crab population monitoring method based on UAV remote sensing imagery described in the above embodiments. The horseshoe crab population monitoring system based on UAV remote sensing imagery includes: a survey quadrat area setting module 1001, a UAV remote sensing image set acquisition module 1002, an orthophoto data acquisition module 1003, and a horseshoe crab population detection result image acquisition module 1004.

[0077] The survey quadrat area setting module 1001 is used to pre-set several transects parallel to the corresponding coastline in the target intertidal zone based on the topographic information and tidal pattern information of the target intertidal zone, and to set up several survey quadrat areas on the transects. The UAV remote sensing image acquisition module 1002 is used to control the UAV to conduct aerial survey flights over each of the survey sample plots within a preset tidal time window to obtain a UAV remote sensing image set, wherein the UAV remote sensing image set includes UAV remote sensing images corresponding to each of the survey sample plots. The orthophoto data acquisition module 1003 is used to preprocess and optimize the UAV remote sensing images in the UAV remote sensing image set to obtain orthophoto data to be detected. The horseshoe crab population detection result image acquisition module 1004 is used to input the orthophoto data into a preset horseshoe crab population identification and detection model to perform target detection and recognition, and obtain the horseshoe crab population detection result image of the target intertidal region, wherein the horseshoe crab population detection result image is marked with the location identifiers of several horseshoe crab targets.

[0078] It should be noted that the above embodiment of the intertidal horseshoe crab population monitoring system based on UAV remote sensing imagery is only used as an example to illustrate the division of the above functional modules when implementing the intertidal horseshoe crab population monitoring method based on UAV remote sensing imagery. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the equipment can be divided into different functional modules to complete all or part of the functions described above.

[0079] Furthermore, the intertidal horseshoe crab population monitoring system based on UAV remote sensing imagery provided in the above embodiments and the intertidal horseshoe crab population monitoring method based on UAV remote sensing imagery in Embodiment 1 belong to the same concept. The implementation process is detailed in the method embodiment, namely Embodiment 1, and will not be repeated here.

[0080] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and this application also intends to include these modifications and variations.

Claims

1. A method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery, characterized in that, Includes the following steps: Based on the topographic and tidal information of the target intertidal zone, several transects parallel to the corresponding coastline are pre-defined in the target intertidal zone, and several survey quadrats are set up on the transects. Within a preset tidal time window, the UAV is controlled to conduct aerial survey flights over each of the survey quadrat areas to obtain a UAV remote sensing image set, wherein the UAV remote sensing image set includes UAV remote sensing images corresponding to each of the survey quadrat areas. The UAV remote sensing images in the aforementioned UAV remote sensing image set are preprocessed and optimized to obtain orthophoto data to be detected. The orthophoto data is input into a preset horseshoe crab population identification and detection model to perform target detection and recognition, thereby obtaining a horseshoe crab population detection result image of the target intertidal region, wherein the horseshoe crab population detection result image is marked with the location identifiers of several horseshoe crab targets.

2. The method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery according to claim 1, characterized in that, The step of pre-setting several transects parallel to the corresponding coastline in the target intertidal zone based on the topographic and tidal information of the target intertidal zone, and setting up several survey quadrats on the transects, includes: Based on the topographic information and tidal pattern information, the frequency distribution of seawater inundation in the target intertidal zone is determined; Based on the seawater inundation frequency distribution, three pre-defined transects are set parallel to the corresponding coastline. The three transects are arranged sequentially in a direction perpendicular to the coastline and correspond to three pre-defined gradient threshold intervals from low to high in the seawater inundation frequency distribution. On each transect line, 3 to 6 survey quadrats with a size of 10 meters × 10 meters are arranged at equal intervals; wherein, on the same transect line, the distance between any two adjacent survey quadrats is 100 meters.

3. The method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery according to claim 2, characterized in that, The step of controlling the UAV to conduct aerial survey flights over each of the survey sample plots within a preset tidal time window to obtain a set of UAV remote sensing images includes: The coordinate information of the surveyed sample plot area was obtained using RTK positioning technology; Based on the coordinate information of the survey sample area, generate UAV aerial survey routes; Within a preset tidal time window, the UAV is controlled to conduct aerial survey flights according to the UAV aerial survey route and collect UAV remote sensing images corresponding to each of the survey sample areas to obtain the UAV remote sensing image set. The preset tidal time window is 0.5 hours after the spring tide recedes; the forward overlap rate of the UAV is 80%; the lateral overlap rate of the UAV is 70%; the flight altitude of the UAV during aerial surveying is 5 to 10 meters; and the ground resolution of the UAV remote sensing images collected by the UAV is not less than 0.5 centimeters.

4. The method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery according to claim 1, characterized in that, The steps of preprocessing and optimizing the UAV remote sensing images in the UAV remote sensing image set to obtain the orthophoto data to be detected include: Image stitching, distortion correction and geographic coordinate alignment are performed on the UAV remote sensing images in the UAV remote sensing image set to generate intermediate orthophoto data. The intermediate orthophoto data is subjected to illumination equalization using a contrast-limited adaptive histogram equalization method to obtain the orthophoto data to be detected. ； Where I represents the intermediate orthophoto image data, clipLimit is a preset contrast limit coefficient set to 2.0, and tileGridSize is the corresponding tile grid size in the illumination equalization process, which is 8 rows × 8 columns. I clahe The orthophoto data to be detected is CLAHE(), which is a preset equalization function.

5. The method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery according to claim 1, characterized in that, The preset horseshoe crab population identification and detection model is a target detection neural network model based on the improved YOLO architecture; The improvements include: the feature extraction network of the horseshoe crab population identification and detection model integrates a high-frequency feature enhancement module for extracting detailed features of the horseshoe crab target and a low-frequency feature enhancement module for extracting the overall semantic features of the horseshoe crab target; the feature fusion network of the horseshoe crab population identification and detection model includes a high-low frequency feature fusion module for fusing features from the high-frequency feature enhancement module and the low-frequency feature enhancement module, and outputting a fused multi-scale feature map; the detection method of the horseshoe crab population identification and detection model is configured as a rotated rectangular bounding box detection; the detection head of the horseshoe crab population identification and detection model is configured to output the parameters of the corresponding rotated rectangular bounding box, the parameters including the center coordinates, width, height, and rotation angle relative to the horizontal axis of the rotated rectangular bounding box.

6. The method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery according to claim 5, characterized in that, The high-frequency feature enhancement module processes the input first feature map in the following manner to output an enhanced high-frequency feature map: The first feature map is subjected to a 1×1 convolution operation, and the resulting convolution is multiplied by a preset spatial attention mechanism weight matrix to obtain the high-frequency feature map: ； in, F high The high-frequency feature map, X in For the first feature map, Conv 1×1 (·) represents the convolution function corresponding to a convolution operation with a 1×1 kernel. S attention The preset spatial attention mechanism weight matrix, F high The high-frequency feature map; The low-frequency feature enhancement module processes the input second feature map in the following manner to output an enhanced low-frequency feature map: The second feature map is subjected to a dilated convolution operation with a dilation rate of 2 and a kernel size of 3×3 to obtain the low-frequency feature map: ； in, F low For the low-frequency feature map, X in ’ For the second feature map, AtrousConv 3×3 (·) represents the dilated convolution function corresponding to a dilated convolution operation with a 3×3 kernel. r =2 represents the void ratio. F low This refers to the low-frequency feature map; The high- and low-frequency feature fusion module fuses the input high-frequency feature map and the low-frequency feature map in the following manner to output the multi-scale feature map: The high-frequency feature map and the low-frequency feature map are concatenated, and the concatenation result is subjected to a preset linear transformation to obtain the multi-scale feature map: ； in, F fusion For the multi-scale feature map, Concat() is the concatenation function corresponding to the concatenation process. W fusion The preset fusion weight matrix, b fusion For the preset bias, F fusion This refers to the multi-scale feature map.

7. The method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery according to claim 5, characterized in that, The horseshoe crab population identification and detection model is obtained by training a preset minimum mixture loss function; the preset minimum mixture loss function is obtained by weighted summation of the rotated box regression loss, the target classification loss, and the confidence loss, and its form is as follows: ； in, L OBB The preset minimum mixture loss function, L reg The regression loss of the rotated frame, L cls The classification loss for the target is... L conf For the confidence loss, α The preset first loss weighting coefficient, β The second loss weighting coefficient is preset. γ The third loss weight coefficient is preset, and α =1.5, β =1.0, γ =0.5; The rotation frame regression loss is calculated using the following formula: ； in, L reg The regression loss of the rotated frame, BB pred For the predicted rotated rectangular bounding box, BB gt For the corresponding labeled true rotated rectangular bounding box, GIoU() is the preset generalized intersection-union ratio (IU) calculation function. λ The preset angle loss weight coefficient and λ =0.8, θ pred For the predicted rotation angle, θ gt This corresponds to the actual rotation angle; The target classification loss is calculated using the following formula: ； in, L cls The classification loss for the target is... α t The preset positive and negative sample weight coefficients and α t =0.25, δ is the preset difficult sample mining coefficient and δ=2, P t The predicted classification probability of the horseshoe crab target by the horseshoe crab population identification and detection model; The confidence loss is calculated using the following formula: ； in, L conf For the confidence loss, y This represents the true confidence level label and takes a value of 1 or 0. σ (·) represents the preset sigmoid activation function. p The prediction confidence level of the horseshoe crab population identification and detection model is given.

8. The method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery according to claim 5, characterized in that, The horseshoe crab population identification and detection model has a species classification sub-network connected to its detection head; the species classification sub-network is used to classify the detected horseshoe crab targets based on the horseshoe crab population detection result image. After the step of obtaining the horseshoe crab population detection result image of the target intertidal region, the intertidal horseshoe crab population monitoring method further includes: Based on the biological morphological information of the Chinese horseshoe crab, the biological feature vector V of the Chinese horseshoe crab is obtained. C : V C =(V C1 ,V C2 ,V C3 ); Among them, V C1 V represents the cephalothorax shape characteristics of the Chinese horseshoe crab. C2 V represents the sharpness of the lateral spines of the horseshoe crab described above. C3 The number of lateral spines of the Chinese horseshoe crab; Based on the morphological information of the horseshoe crab, the biological feature vector V of the horseshoe crab is obtained. R : V R =(V R1 ,V R2 ,V R3 ); Among them, V R1 V represents the cephalothorax shape characteristics of the horseshoe crab. R2 V represents the sharpness of the lateral spines of the horseshoe crab. R3 The number of lateral spines of the horseshoe crab; The horseshoe crab population detection result image is input into the species classification sub-network, and the species classification sub-network is driven to process the horseshoe crab population detection result image according to the biofeature vectors of the Chinese horseshoe crab and the roundtail horseshoe crab to obtain the species classification probability corresponding to the detected horseshoe crab targets: P spec =(PC,PR); Among them, P spec The probability of the species classification corresponding to the detected horseshoe crab target is PC, which represents the probability that the corresponding horseshoe crab target is the Chinese horseshoe crab, and PR represents the probability that the corresponding horseshoe crab target is the roundtail horseshoe crab. Based on the species classification probability, the species of the corresponding horseshoe crab target is determined. If the PC is greater than 0.5, the horseshoe crab target is determined to be the Chinese horseshoe crab. If the PR is greater than 0.5, the horseshoe crab target is determined to be the roundtail horseshoe crab.

9. The method for monitoring intertidal horseshoe crab populations based on UAV remote sensing imagery according to claim 5, characterized in that, After obtaining the horseshoe crab population detection result image of the target intertidal region, the intertidal horseshoe crab population monitoring method further includes the step of acquiring the individual biological parameters of the corresponding horseshoe crab target: For the horseshoe crab target in the horseshoe crab population detection result image, based on its corresponding rotated rectangular bounding box, the pixel width of its forebody armor is extracted using a preset edge detection algorithm: ； The left and right edge pixel coordinates of the horseshoe crab target's forebody are respectively ( x 1, y 1) and ( x 2, y 2) θ The rotation angle of the rotating rectangular bounding box. W p The pixel width; Based on the ground resolution of the orthophoto data and the pixel width, the actual physical width of the horseshoe crab's forebody armor is calculated: ； in, W real The actual physical width of the forebody armor of the horseshoe crab target. GSD The ground resolution of the orthophoto data is given. k This is a preset unit conversion factor; If the horseshoe crab's protocarp is occluded in the horseshoe crab population detection result image, the width of the horseshoe crab's crawl marks is obtained, and the actual physical width of the horseshoe crab's protocarp is calculated using a preset correction formula. W real =1.2× W trace +0.8； in, W trace The width of the crawl trail of the horseshoe crab target; The actual physical width is input into a preset age inversion model to obtain the predicted age of the horseshoe crab target: ； in, a and b These are the preset regression coefficients of the first model and the regression coefficients of the second model, respectively. a =4.263, b =-7.892, N pred The predicted age of the horseshoe crab target.

10. A monitoring system for intertidal horseshoe crab populations based on UAV remote sensing imagery, characterized in that, include: The survey quadrat area setting module is used to pre-set several transects parallel to the corresponding coastline in the target intertidal zone based on the topographic information and tidal pattern information of the target intertidal zone, and to set up several survey quadrat areas on the transects. The UAV remote sensing image acquisition module is used to control the UAV to conduct aerial survey flights over each of the survey sample plot areas within a preset tidal time window to obtain a UAV remote sensing image set, wherein the UAV remote sensing image set includes UAV remote sensing images corresponding to each of the survey sample plot areas. The orthophoto data acquisition module is used to preprocess and optimize the UAV remote sensing images in the UAV remote sensing image set to obtain the orthophoto data to be detected. The horseshoe crab population detection result image acquisition module is used to input the orthophoto data into a preset horseshoe crab population identification and detection model to perform target detection and recognition, and obtain the horseshoe crab population detection result image of the target intertidal region, wherein the horseshoe crab population detection result image is marked with the location identifiers of several horseshoe crab targets.

Images